Low-Dimensional Subject Representation-Based Transfer Learning in EEG Decoding

Abstract

Recently, the advances in passive brain-computer interfaces (BCIs) based on electroencephalogram (EEG) have shed light on real-world neuromonitoring technologies. However, human variability in the EEG activities hinders the development of practical applications of EEG-based BCI. To tackle this problem, many transfer-learning techniques perform supervised calibration. This kind of calibration approach requires taskrelevant data, which is impractical in real-life scenarios such as drowsiness during driving. This study presents a transferlearning framework for EEG decoding based on the lowdimensional representations of subjects learned from the pre-trial EEG. Tensor decomposition was applied to the pre-trial EEG of subjects to extract the underlying characteristics in subject, spatial and spectral domains. Then, the proposed framework assessed the characteristics to obtain the low-dimensional subject representations such that the subjects with similar brain dynamics can be identified. This method can leverage the existing data from other users and a small number of data from a rapid, nontask, unsupervised calibration from a new user to build an accurate BCI. Our results demonstrated that, in terms of prediction accuracy, the proposed low-dimensional subject representationbased transfer learning (LDSR-TL) framework outperformed the random selection and the Riemannian manifold approach in cognitive-state tracking, while requiring fewer training data. The results can greatly improve the practicability and usability of EEG-based BCI in the real world.

Multi-Armed Bandits in Brain-Computer Interfaces

“…Using data from previous sessions or persons and using transfer learning to adapt the old data to the current session is one solution (Lotte, 2015 ). To find relevant data, one can among other approaches use tensor decomposition (Jeng et al, 2021 ), Riemannian geometry (Khazem et al, 2021 ), or a generic machine learning model (Jin et al, 2020 ). In Gutiérrez et al ( 2017 ), they use the classic MAB problem to find clusters of data in a big medical data set which increases the classification accuracy.…”

Comparison of Domain Selection Methods for Multi-Source Manifold Feature Transfer Learning in Electroencephalogram Classification

“…The primary focus of domain selection is to minimize EEG variations between the TD and SDs, mainly to minimize the effect of NT, which has a significant impact on classification performance across individual TDs [14,15]. ROD iterates through domains to identify less-beneficial sources, and, in each iteration, a single less-beneficial domain is discarded, while the remaining domains are considered SDs, and a single domain is considered TD when transfer mapping is applied.…”

Multi-Source Decentralized Transfer for Privacy-Preserving BCIs

“…Zhang and Wu [17] proposed a manifold embedded knowledge transfer approach for MI, which combines data alignment and feature adaptation in the tangent space. Jeng et al [18] introduced a low-dimensional representation-based TL framework for MI decoding based on tensor decomposition. Xu et al [15] proposed an instance-based selective TL approach for MI in the Riemannian tangent space, which utilizes labeled samples from the source and target subjects.…”

“…Transfer learning (TL) [11] is a promising approach to alleviate this problem. Various TL approaches have been proposed for BCI in the last decade, e.g., adaptive CSP [12], data alignment [13], [14], instance-based TL [15], [16], feature-based TL [17], [18], and deep TL [19]. For aBCIs, existing TL approaches mainly include feature-based TL [20] and adversarial-based deep TL [21], [22].…”